Réf. Böhm & al. 2001 - A

Référence bibliographique complète
BÖHM R., AUER I., BRUNETTI M., MAUGERI M., NANNI T., SCHÖNER W. Regional temperature variability in the European Alps: 1760-1998 from homogenized instrumental time series. Int. J. Climatol., 2001, Vol. 21, p. 1779-1801.

Abstract: This paper investigates temperature variability in the Alps and their surroundings based on 97 instrumental series of monthly mean temperatures. Each of the original series had breaks (an average of five per series) and the mean of all series was systematically biased by non-climatic noise. This noise has subdued the long-term amplitude of the temperature evolution in the region by 0.5 K. The relatively high spatial resolution of the data enabled a regionalization within the study area of 680 000 km2 into six sub-regions based on principal component analysis of the monthly series. Long-term temperature evolution proved to be highly similar across the region—thus making a mean series (averaged over all 97 single series) representative of the study area. Trend analysis (based on progressive forward and backward Mann–Kendall statistics and on progressive analysis of linear regression coefficients) was performed on seasonal and annual series. The results diverge from those of global datasets. This is mainly due to the extension of the 240-year Alpine dataset by 100 years prior to the mid-19th century, and also due to the advantages of a dense and homogenized regional dataset. The long-term features include an initial decrease of the annual and seasonal series to a minimum followed by a positive trend until 1998. The minima are 1890 for the entire year and winter, 1840 for spring and 1920 for summer and autumn, respectively. The initial decreasing trend is more evident in spring and summer, less in autumn and smallest in winter. The mean annual temperature increase since 1890 in the Alps is 1.1 K, which is twice as much as the 0.55 K in the respective grid boxes of the most frequently used global dataset of the Climatic Research Unit (CRU), University of East Anglia.

Mots-clés
Gridded dataset, homogeneity, instrumental period, regionalization, temperature time series, trends.

Organismes / Contact
Central Institute for Meteorology and Geodynamics, Hohe Warte 38, A -1190 Vienna , Austria. reinhard.boehm@zamg.ac.at
Istituto ISAO-CNR, Via Godetti, 101, I-40129 Bologna, Italy.
Istituto di Fisica Generale Applicata, Universita di Milano, Via Brera, 28, I-20121 Milano, Italy.

(1) - Paramètre(s) atmosphérique(s) modifié(s)
(2) - Elément(s) du milieu impacté(s)
(3) - Type(s) d'aléa impacté(s)
(3) - Sous-type(s) d'aléa
Temperature      

Pays / Zone
Massif / Secteur
Site(s) d'étude
Exposition
Altitude
Période(s) d'observation
France, Germany, Switzerland, Italy, Austria, Slovenia, Croatia, Hungary and Slovakia Alps and their surroundings 97 meteorological stations     1760-1998

(1) - Modifications des paramètres atmosphériques
Reconstitutions
 
Observations
The following regions have been identified as homogeneous temperature areas:
- High-level Alpine areas: These include the scattered areas lying above 1500 m a.s.l. This station cluster is clearly separated from the low-level areas. Most of its group members are located at remote places on mountain peaks;
- The East: A region of mainly continental features, framing the Alps in the north, the east and the south-east;
- The West: A region of maritime influences from the Atlantic and the western Mediterranean. The border to region East is a zone of gentle transition rather than a clear line;
- The South: A region covering the Adriatic coast and the northern part of the Italian peninsula;
- The Po plain: A region covering most of the northern continental part of Italy. It has definite borders in the West and North (towards the Alps) in contrast with the situation in the East, which is characterized by a gentle transition into the Adriatic cluster;
- The central Alpine low-level areas: A region that concentrates on valley stations in the central part of the Alps where they have their maximum north–south extension plus parts of the eastern Alps.

The high similarity between the ALPCLIM regional averages gives a preliminary representation of the whole area by means of only one series: the mean for all ALPCLIM sites (hereafter MA). The annual data show a start from lower values before 1785, followed by two relative maxima in the 1790s and the 1820s, interrupted by a sudden cold event in the 1810s. After the 1820s there is a gradual trend toward two minima in the 1840s and 1850s and the 1880s and 1890s with a relative maximum in the 1860s and early 1870s. The whole 20th century is characterized by rising temperatures toward a first maximum at 1950 and a second in the 1990s, which is the main maximum of the 240 ALPCLIM years. The analysis of the seasonal series provides evidence that, beside some common features, the temperature evolution shows strong seasonal differences. Concentration on the long-term features alone, annual and seasonal MA series initially decrease to a minimum, which is then followed by a positive trend until 1998. The location of the minima is different however: 1890 for the entire year and winter, 1840 for spring and 1920 for summer and autumn. The initial decreasing trend is more evident in spring and summer, less in autumn and least in winter. The mean annual temperature increase since 1890 in the Alps is 1.1 K. The ALPCLIM series has a 0.5 K per 100 years warming trend in summer and a 1.1–1.3 K per 100 years warming trend in winter.

The length of the ALPCLIM series enables an analysis of data covering the pre-industrial period, which is frequently used for comparison with the warmer 20th century climate. This changes the results of the trend analysis because the early period is not characterized by low temperatures but shows high temperatures, especially in spring and summer. As a consequence, the period of significantly positive trends (of the entire 240 years) is rather short in the Alps (starting around 1970 in winter and autumn and in the 1980s for the annual mean). For spring there has been a recent evolution to positive trends but as yet these are not significant. Summers are characterized by significantly negative trends for more than one and a half centuries and also during the entire 240 years the summer trend is still significantly negative.
Modélisations
 
Hypothèses
 

Informations complémentaires (données utilisées, méthode, scénarios, etc.)
120 single series were collected, all based on monthly means, most of them being at least centennial-length series, the longest series starting in the 1750s. The ALPCLIM temperature dataset was constructed from different sources of several countries (France, Germany, Switzerland, Italy, Austria, Slovenia, Croatia, Hungary and Slovakia). All series, original and pre-homogenized, were re-analysed for inhomogeneities based on the following system: homogeneity testing, adjusting and gap-closing was performed in 13 regional sub-groups of ten series using the MASH-test of Szentimrey (1999) and the HOCLIS procedure (Auer et al., 1999). HOCLIS test each series against other series in sub-groups of ten series. The break signals of one series against all others are then collected in a decision matrix and the breaks are assigned to the single series according to probability. One additional sub-group was created for the pre-1820 parts of the ten bicentennial series. The last homogenizing groups consisted of two high-elevation sites with series above 1500 m a.s.l. for the western and for the eastern Alps.

A final total of 97 single series proved to be homogenizable. The mean station distance for the 682 000 km2 of the ALPCLIM region is 80 km. Three homogenized series start as early as 1760, ten earlier than 1801 and there was 25 series in 1850. The greatest increase in station density occurred in the 1860s. The average of the adjusted series reveals a general increasing trend of approximately 0.5 K for the 150 years from the mid-19th to the end of the 20th century. The reasons for this systematic trend can be found in a mixture of specific national items and some general evolutions in the study area.

The first step in analysing the data was the transformation of the 97 homogenized series into anomaly series. Anomalies were calculated on a monthly basis, subtracting the monthly mean for the 1901–1988 period (hereafter referred to as the 20th century mean). A correlation analysis was then performed. The results show that the monthly temperature anomalies have a high spatial correlation over the ALPCLIM region and that the difference in height affects the correlation more than the horizontal distance.

Globally, the results of the correlation analysis suggest that most of the signal in the ALPCLIM monthly temperature anomalies can be captured by a small number of regional average series. Following this result, the authors clustered their series into homogeneous temperature areas by means of principal component analysis (PCA). PCA allows the identification of a small number of variables known as principal components (PCs), which are linear functions of the original data, which maximize the amount of their explained variance. The technique can be applied both to correlation and covariance matrixes. The same regions identified from the PCA of the complete series have also been identified in the seasonal PCA. The main difference is between the summer and winter analyses. During summer the correlation extends over larger areas both in the horizontal plane and along the vertical, whereas in winter the local effects are more important.

After clustering the stations, the authors constructed mean regional anomalies series performing the arithmetic mean of the series located by PCA in each of these six regions. Beside the high correlations identified between the regional mean series and their corresponding cluster members, there are also significant correlations between the different regional mean series. On average, the overall sites have very high correlation with all low-level regional series and also good correlation with the high level series.

The first step in trend analysis was the calculation of seasonal and annual average regional series. Annual values correspond to the period from December to November and are dated by the year in which January is included. Thus, the annual mean covers the same months as the seasonal means, where winter values refer to the December–February interval, spring to March–May, summer to June–August and autumn to September–November. All values are anomalies from the 20th century mean. It is evident that the regional temperature anomalies are highly correlated, not only on an annual time scale but also on a secular time scale with a range of the filtered curves generally less than 0.5 K. Prior to 1850, the range gently increases, which might be due to the much smaller station density in the early parts of the series.

In order to quantify these results, the slopes of the trends were calculated by least square linear fitting and their confidences were estimated by means of the non-parametric Mann–Kendall test (Sneyers, 1990). Globally, trend analysis shows that the significance and the slope of the trends depend strictly on the selected periods, with generally higher values in winter than in summer. However, for a given period and season, the results are highly similar for all the different ALPCIM regions.

(2) - Effets du changement climatique sur le milieu naturel
Reconstitutions
 
Observations
 
Modélisations
 
Hypothèses
 

Sensibilité du milieu à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)
   

(3) - Effets du changement climatique sur l'aléa
Reconstitutions
 
Observations
 
Modélisations
 
Hypothèses
 

Paramètre de l'aléa
Sensibilité des paramètres de l'aléa à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)
 
 

(4) - Remarques générales
 

(5) - Syntèses et préconisations
 

Références citées :

Auer I, Böhm R, Schöner W, Hagen M. 1999. ALOCLIM—Austrian–Central European long-term climate—creation of a multiple homogenised long-term climate data-set. In Proceedings of the 2 nd Seminar for Homogenisation of Surface Climatological Data . Budapest, 9–13 November 1998, WCDMP-No. 41, WMO-TD No. 962; 47–71.

Sneyers R. 1990. On the statistical analysis of series of observations. WMO-TN 143.

Szentimrey T. 1999. Multiple analysis of series for homogenisation (MASH). In Proceedings of the 2 nd Seminar for Homogenisation of Surface Climatological Data . Budapest, 9–13 November 1998, WCDMP-No. 41, WMO-TD No. 962; 27–46.